An AI-Driven Framework for Early Detection of Stuck Pipe Events in Drilling Operations

Stuck pipe events are a major cause of non-productive time (NPT) and increased operational costs in drilling operations. This study presents an advanced AI-driven framework for early detection of stuck pipe events, integrating fuzzy logic, precursor-based feature engineering, and machine learning. The framework processes high-frequency drilling data, including weight on bit (WOB), standpipe pressure (SPP), and torque (TQA), to provide a probabilistic risk assessment. Fuzzy logic generates interpretable risk levels, while precursor-based features identify specific event triggers such as stick-slip events, low SPPA-SPMT correlation, and abrupt trajectory variations. A gradient-boosted decision tree model, trained on historical data using a sliding window approach, predicts the probability of stuck pipe events, enabling proactive decision-making.

Tested on a global dataset of 350+ wells, the framework demonstrates strong performance in reducing false alarms and providing actionable insights. The probabilistic approach, combined with precursor analysis and fuzzy logic, offers a nuanced risk gradient, allowing drilling teams to prioritize actions based on the severity of risk trends. This integrated solution improves operational efficiency, reduces NPT, and lowers drilling costs. Future work will focus on enhancing model generalizability across diverse drilling environments and refining visualization tools for improved user interaction.